Machine Vision
	The field of image processing has expanded over the years to include many sub-fields, each with its own terminology and commercial marketplace. This sometimes confusing array of overlapping technologies includes the terms "image analysis", "image understanding", "machine vision", "image enhancement and restoration", and other more specific applications such as "biometrics", "facial recognition", "motion tracking", and others. Here is a brief description of these terms to help differentiate them. Image Processing Nowadays, the term "image processing" can refer to the general concept of manipulating, displaying and storing visual images, or it can refer to the more narrow set of techniques for processing image data with low-level pixel-based algorithms such as filtering, smoothing, convolution, and histogram equalization. These techniques typically take an input image, process each pixel sequentially, and produce an output image of the same size. The output image has some desirable improvement over the input image such as improved noise, smoothness, better viewability, etc. This narrower definition is sometimes referred to as "image conditioning". Image Analysis and Image Understanding Image analysis or image understanding takes image processing one step further and attempts to measure some variable or set of variables from an image. The variable can be a yes/no decision or it can be a continuous variable that corresponds to a physical measurement or the probability of the image containing some feature of interest. For example, one image analysis algorithm could determine if the input image contains a person's face. Another could provide an estimate of the volume of a patient's left ventricle from a series of CT scans. These higher-level algorithms would depend on applying a variety of low-level image processing functions. Machine Vision and Robotic Vision Machine vision can be thought of as an application of image analysis in the manufacturing or inspection fields. A typical machine vision system involves one or more cameras viewing an object being manufactured to direct the manufacturing process. Or the system inspects manufactured objects for quality control. There is a large market for general purpose machine vision systems that are very flexible and that can be easily programmed to detect and measure a wide variety of objects. These systems also rely on low-level image processing algorithms, typically to measure the dimensions of objects by detecting their edges. Robotics systems sometimes have a machine vision component. Image Enhancement and Restoration Image enhancement and restoration refers to a set of image processing algorithms that attempts to improve the view ability or recognition of an image that has been degraded with noise, movement, blurring or other artifacts. Depending on the problem area, these algorithms rely on image filtering, "histogramming" and other pixel-based techniques. Image Compression and Transmission Image compression refers to the reduction in the storage size of an image by removing "redundant data" in the image. This is done by applying transform-based algorithms such as the Fourier transform, the Cosine transform (the basis of JPEG compression) and wavelet transforms. These algorithms are starting to be incorporated in storage hardware making their application transparent to the user.